The Uncarved Block

I was quite interested to read about the idea of "learning in public", a strategy that involves sharing notes, plans and progress online to benefit others and reinforce your own motivation and commitment. Given my own blog could use some updates is available and written using markdown, like my notes in obsidian it is a natural and easy fit, so I'm going to experiment with making more and more of my notes available online via this mechanism.

You can find my notes as I go through this process online at https://publish.obsidian.md/uncarved. A good place to start might be my note on self-directed learning.

This is a static website hosted on AWS using S3 and Cloudfront. The site is generated using zola, which is a very fast and no-nonsense static site generator written in rust.

I also use AWS to host DNS for my sites and use fastmail for mail hosting, which means mail just works and I don't have to spend any time figuring out why it's broken or why upstream sites aren't accepting my mail. For some time now people hosting their own mail and dns have been at a big disadvantage due to the way in which large network providers have chosen to conduct the war on spam.

My pages are written using the neovim editor, and I use a Planck ortholinear dvorak keyboard.

Note: Amusingly almost all of the above has changed since I wrote the first time I made this page, so I've kept the old one around for comparison.

They checked 3 times that I wanted lemon (not cucumber) in my second $38 Hendricks and tonic even though I very obviously had lemon in my first one. Then they brought cucumber anyway. The obscene price and the cucumber and lemon thing are par for the course here but they are now trying to charge me a second time for the drink I already paid for. I give them the receipt showing I paid for that one.

Settling the bill for this drink has already taken over 10mins with three staff squinting at the machine, my card and the receipt. And I'm not done yet.

During the process I confirm my name twice while they stare in amazement at my credit card. Probably not everyone has credit which stretches to two drinks in this place.

The bar is very full. I can only assume everyone else is stuck here still trying to settle up for drinks they bought days or even weeks ago. Lots of people seem to have Hendricks and tonic (with cucumber) which is a fair indication I'm onto something.

I'm deeply alone, on the last day of a business trip. My friends said they would meet me here then all obviously ended up somewhere else. There is no reception so we are unlikely to connect. I will not see many of them for months. I used to love working with them, now see them very seldom and the thought fills me with a profound sense of sadness.

One bar person comes over to explain the confusion. "You had 2 Hendricks and tonic, one with lemon and one with cucumber."

I finish my drink, get my coat and head out into the cold.

Often you will see in computer books and articles a pattern where a function is applied to some but not all of it's required arguments, resulting in a function of fewer arguments. In python, this looks like this (from PEP 309:

def papply(fn, *cargs, **ckwargs):
    def call_fn(*fargs, **fkwargs):
        d = ckwargs.copy()
        d.update(fkwargs)
        return fn(*(cargs + fargs), **d)
    return call_fn

This is called "partial function application" - it returns a function which acts like the function you pass in but which takes fewer arguments, the others having been "bound in". The author of this code, however, had the (very common) misconception that this is currying, and called his function "curry" as a result. I shared this misconception for some time, and thought that currying and partial application were the same thing. In fact they are to certain extent opposites.

Where partial application takes a function and from it builds a function which takes fewer arguments, currying builds functions which take multiple arguments by composition of functions which each take a single argument. Thus we curry in python like this:

def addN(n):
    return lambda x: x + n

def plus(a, b):
    addA=addN(a)
    return addA(b)

Now why would we ever want to do that? Well, in some pure functional languages this is exactly how functions with multiple arguments are built up. In ocaml, a function which takes two ints and returns a float is actually a function which takes an int and returns a function which takes an int and returns a float. In this world, partial application just happens without any extra code:

% rlwrap ocaml
    Objective Caml version 3.09.3

# let add a b=a+b;;
val add : int -> int -> int = <fun>

So the type of add is a function which takes an int and returns a function which takes an int and returns an int.

# let add2=add 2;;
val add2 : int -> int = <fun>

add is a curried function, so here we can partially apply by just calling with a single arg- it returns the function that takes the other arg and returns the result.

# add2 34;;
- : int = 36

...and we can call add2 with a single argument as you would expect. Because ocaml curries add for us, the function has been partially applied. It's interesting to note that in ocaml if you label your function arguments, they can be partially applied in any order.

I shouldn't have to explain this to you but...

“That isn't a falafel sandwich. It's some hummus and some pieces of pitta bread."

"Yeah, but you ordered the falafel sandwich."

"I did. That's not a falafel sandwich though."

"See, it says here. You ordered the falafel sandwich."

"Yeah. That doesn't have falafel. And it's not a sandwich."

She looks from the printout to the plate a few times in disbelief. A woman at the next table is trying to attract her attention to get the hummus she obviously ordered, but my server is not the type to be so easily swayed. This isn’t the first technical hitch in the delivery of my meal either. Before the absence of sandwich I was brought a bowl of soup but no spoon. After asking three times they handed over a teaspoon. Now there is a brief impasse after which the non-sandwich is taken away. Time passes. I’m sitting in a restaurant/bar in Liberty International Airport in Newark, New Jersey. Which, you know. New Jersey. But at least I’m leaving.

The setup in this particular place is that there is an iPad at each table which you use to order and pay. I have a headache and they're playing loud Cuban music. The food is not at all cheap, but the whole iPad schtick means they know precisely who ordered what and presumably is so they can get away with not paying their staff- hence the prodigiously awful service.

It’s spring and there’s enough snow falling for me not to be able to really see the runway even though I’m right in front of a window looking directly at it. Planes nearby have snow gathering in drifts on their wings. Other planes seem to be landing and taking off, which is good because I don’t really want to stay in New Jersey.

"Hi. I'm still waiting for a falafel sandwich."

"I thought it was brought to you. "

"No, you brought hummus. And I told you it wasn't a falafel sandwich. Then you took it away."

"I thought it was brought to you. Lemme check."

"You can see that nothing was brought to me. I don’t have any plates. And my cutlery is unused."

"Lemme check, I thought it was brought to you."

She moves diagonally by one table, remaining in my field of view and starts ostentatiously wiping down a table with one hand while texting on her phone with the other. She wants to make it as clear as possible by her actions that she has OTHER THINGS TO DO and is most certainly not in any way checking for my sandwich. I redirect some wine that was ordered by the guy next to me.

Eventually my food comes without her “help". My knife is made of bendy plastic and when I use it to push a piece of lettuce onto my fork it snaps in half for airline safety reasons.

Here's how to make delicious smooth cold-brew coffee:

1. You need a cafetière (French press for my American friends) and a drip filter (like if you were a hipster making a pourover). A chemex works great for this.
2. Ok, now grind the coffee fine as though you're making espresso (that's not what you'd use for a regular cafetière coffee I know). I use about 68g of coffee for 800ml cafetière. Your mileage may vary, but err on the side of more coffee rather than too little. 1.5 - 2 tablespoons per cup of coffee is I think the rule for this sort of thing.
3. Put the ground coffee in the cafetière and pour on regular cold tap water. You might want to use filtered water if your tap water doesn't taste good enough to drink on its own.
4. Give it a quick stir, wait ten minutes & give it another quick stir. Don't put the top on the cafetière, just put some clingfilm over the top and put the whole thing in the fridge.
5. Wait 24 hours.
6. Take the whole thing out of the fridge, remove the clingfilm, give it a quick stir, put the top on the cafetière and press down just like you normally would if you were using the thing to make hot coffee.
7. You don't want sludge at the bottom of your cup, so pour it through the drip filter into whatever container you want to store the coffee in (I use a water bottle normally) and keep it in your fridge.

You now have a lovely bottle of the coffee equivalent of blue meth in your fridge. Be warned this mix is strong. I like to drink it in two different ways:

1. Neat in an espresso cup as a replacement for my first espresso of the day.
2. Over lots of ice diluted about 2 parts cold brew to 1 part water if I want a coffee to take with me.

Enjoy!

Short version

1. Make coffee in a cafetière/French press using cold tap water and a fine grind for the beans (not coarse as would be normal)
2. Put it in the fridge to brew for 24 hours
3. Plunge it, then run it through a drip filter

I trained to postgraduate level as a Jazz bass player, but worked as a programmer to pay for my music postgrad and somehow ended up in a career in computer science. After various roles as a permanent employee and contractor I became the IT director of a dot com company. I then joined Goldman Sachs because the role I was offered sounded very interesting. I stayed there for 8 years, working initially on infrastructure for distributed pricing and risk, then in what would now be called devops, and then on a number of different trading desks. It was indeed a formative experience and I learned a lot although it was very challenging at times.

I eventually left GS and joined Palantir Technologies, a silicon valley software company that focusses on data analysis. Palantir attracts a lot of attention from conspiracy theorists due to work they believe it does for the US government however I worked on the commercial side, focussing on detecting rogue traders. This led to a joint venture they formed with Credit Suisse, called Signac, which I was fortunate enough to be co-head of. Unfortunately due to changes in the regulatory landscape the founding rationale for this startup went away and Credit Suisse and Palantir decided to dissolve this partnership.

I then joined OakNorth as the CIO of their platform business, which uses data science, tech and credit expertise to transform lending to medium-sized and growth businesses. It's been a very exciting journey so far.

I'm also on the panel of advisors of Antler, a hybrid incubator/vc and am an advisor to a startup called Kamayi. I'm first author on 2 patents in the field of fraud detection and wrote a chapter of "The Regtech Book" published by Wiley Press along with a number of other articles. I'm on the editorial board of the Journal of Digital Banking.

I've spoken at a number of large tech and fintech events both in person and on-line, including AWS Re:Invent, Money 2020, the MAS Fintech festival, the Inovate Finance Global Summit, Wired:Smarter, CogX and many others.

You can see my linkedin profile, and invite me to your network if we know each other. There are quite a few Sean Hunters though so I might not be the one you're thinking of. You can email me if you like and I will try to respond although I can't promise to be too prompt.

My interest in the I Ching began when I was a music student. Studing John Cage, I learned that he was heavily influenced by the I Ching and used hexagrams as an aid to composition of aleatoric music. I bought a copy of the Richard Wilhelm translation and learned to cast hexagrams via the coin oracle. I revived this interest when I studied Tai Chi Chuan.

Every visitor to the front page of this website is greeted by a pair of I-Ching hexagrams, which are generated by this small python program. The distribution of results in the I-Ching is not uniform, and this page gives the same distribution of lines as the Yarrow Stalk oracle. The second hexagram is the first but with "moving lines" inverted in the traditional fashion. This website uses web.py, and I incorporate the I-Ching reading into the Cheetah template for the homepage. The relevant snippet of template code looks like this:

#import ching
#set $hexes = $ching.get_hexagram_pair
#set $hex1 = int($ching.get_hexagram_number($hexes[))][0]
#set $hex2 = int($ching.get_hexagram_number($hexes[))][1]
#set $name1 = $ching.get_hexagram_name($hexes[)][0]
#set $name2 = $ching.get_hexagram_name($hexes[)][1]
#echo '<img id="hex1" alt="%d: %s" src="/static/images/iching/Iching
-hexagram-%02d.png" />' % (hex1,name1 ,hex1) #
#echo '<img id="hex2" alt="%d: %s" src="/static/images/iching/Iching
-hexagram-%02d.png" />' % (hex2,name2 ,hex2) #

...and the lovely hexagram graphics are in the public domain. I downloaded them from the wikipedia.